Image-derived Metrics Research Articles

Six-minute, in vivo MRI quantification of proximal femur trabecular bone 3D microstructure.

Assessment of proximal femur trabecular bone microstructure in vivo by magnetic resonance imaging has recently been validated for acquiring information independent of bone mineral density in osteoporotic patients. However, the requisite signal-to-noise ratio (SNR) and resolution for interrogation of the trabecular microstructure at this anatomical location prolongs the scan duration and renders the imaging protocol clinically infeasible. Parallel imaging and compressed sensing (PICS) techniques can reduce the scan duration of the imaging protocol without substantially compromising image quality. The present work investigates the limits of acceleration for a commonly used PICS technique, ℓ1-ESPIRiT, for the purpose of quantifying measures of trabecular bone microarchitecture. Based on a desired error tolerance, a six-minute, prospectively accelerated variant of the imaging protocol was developed and assessed for intersession reproducibility and agreement with the longer reference scan. To investigate the limits of acceleration for MRI-based trabecular bone quantification by parallel imaging and compressed sensing reconstruction, and to develop a prototypical imaging protocol for assessing the proximal femur microstructure in a clinically practical scan time. Healthy participants (n=11) were scanned by a 3D balanced steady-state free precession (bSSFP) sequence satisfying the Nyquist criterion with a scan duration of about 18min. The raw data were retrospectively undersampled and reconstructed to mimic various acceleration factors ranging from 2 to 6. Trabecular volumes-of-interest in four major femoral regions (greater trochanter, intertrochanteric region, femoral neck, and femoral head) were analyzed and six relevant measures of trabecular bone microarchitecture (bone volume fraction, surface-to-curve ratio, erosion index, elastic modulus, trabecular thickness, plates-to-rods ratio) were obtained for images of all accelerations. To assess agreement, median percent error and intraclass correlation coefficients (ICCs) were computed using the fully-sampled data as reference. Based on this analysis, a prospectively 3-fold accelerated sequence with a duration of about 6min was developed and the analysis was repeated. A prospective acceleration factor of 3 demonstrated comparable performance in reproducibility and absolute agreement to the fully-sampled scan. The median CoV over all image-derived metrics was generally <6% and ICCs >0.70. Also, measurements from prospectively 3-fold accelerated scans demonstrated in general median percent errors of <7% and ICCs >0.70. The present work proposes a method to make in vivo quantitative assessment of proximal femur trabecular microstructure with a clinically practical scan duration of about 6min.

VistoSeg: Processing utilities for high-resolution images for spatially resolved transcriptomics data.

Spatially resolved transcriptomics (SRT) is a growing field that links gene expression to anatomical context. SRT approaches that use next-generation sequencing (NGS) combine RNA sequencing with histological or fluorescent imaging to generate spatial maps of gene expression in intact tissue sections. These technologies directly couple gene expression measurements with high-resolution histological or immunofluorescent images that contain rich morphological information about the tissue under study. While broad access to NGS-based spatial transcriptomic technology is now commercially available through the Visium platform from the vendor 10× Genomics, computational tools for extracting image-derived metrics for integration with gene expression data remain limited. We developed VistoSeg as a MATLAB pipeline to process, analyze and interactively visualize the high-resolution images generated in the Visium platform. VistoSeg outputs can be easily integrated with accompanying transcriptomic data to facilitate downstream analyses in common programing languages including R and Python. VistoSeg provides user-friendly tools for integrating image-derived metrics from histological and immunofluorescent images with spatially resolved gene expression data. Integration of this data enhances the ability to understand the transcriptional landscape within tissue architecture. VistoSeg is freely available at http://research.libd.org/VistoSeg/.

Task-based functional MRI challenges in clinical neuroscience: Choice of the best head motion correction approach in multiple sclerosis.

Functional MRI (fMRI) is commonly used for understanding brain organization and connectivity abnormalities in neurological conditions, and in particular in multiple sclerosis (MS). However, head motion degrades fMRI data quality and influences all image-derived metrics. Persistent controversies regarding the best correction strategy motivates a systematic comparison, including methods such as scrubbing and volume interpolation, to find optimal correction models, particularly in studies with clinical populations prone to characterize by high motion. Moreover, strategies for correction of motion effects gain more relevance in task-based designs, which are less explored compared to resting-state, have usually lower sample sizes, and may have a crucial role in describing the functioning of the brain and highlighting specific connectivity changes. We acquired fMRI data from 17 early MS patients and 14 matched healthy controls (HC) during performance of a visual task, characterized motion in both groups, and quantitatively compared the most used and easy to implement methods for correction of motion effects. We compared task-activation metrics obtained from: (i) models containing 6 or 24 motion parameters (MPs) as nuisance regressors; (ii) models containing nuisance regressors for 6 or 24 MPs and motion outliers (scrubbing) detected with Framewise Displacement or Derivative or root mean square VARiance over voxelS; and (iii) models with 6 or 24 MPs and motion outliers corrected through volume interpolation. To our knowledge, volume interpolation has not been systematically compared with scrubbing, nor investigated in task fMRI clinical studies in MS. No differences in motion were found between groups, suggesting that recently diagnosed MS patients may not present problematic motion. In general, models with 6 MPs perform better than models with 24 MPs, suggesting the 6 MPs as the best trade-off between correction of motion effects and preservation of valuable information. Parsimonious models with 6 MPs and volume interpolation were the best combination for correcting motion in both groups, surpassing the scrubbing methods. A joint analysis regardless of the group further highlighted the value of volume interpolation. Volume interpolation of motion outliers is an easy to implement technique, which may be an alternative to other methods and may improve the accuracy of fMRI analyses, crucially in clinical studies in MS and other neurological populations.

Deep learning-guided attenuation correction in the image domain for myocardial perfusion SPECT imaging

Abstract We investigate the accuracy of direct attenuation correction (AC) in the image domain for myocardial perfusion SPECT (single-photon emission computed tomography) imaging (MPI-SPECT) using residual (ResNet) and UNet deep convolutional neural networks. MPI-SPECT 99mTc-sestamibi images of 99 patients were retrospectively included. UNet and ResNet networks were trained using non-attenuation-corrected SPECT images as input, whereas CT-based attenuation-corrected (CT-AC) SPECT images served as reference. Chang’s calculated AC approach considering a uniform attenuation coefficient within the body contour was also implemented. Clinical and quantitative evaluations of the proposed methods were performed considering SPECT CT-AC images of 19 subjects (external validation set) as reference. Image-derived metrics, including the voxel-wise mean error (ME), mean absolute error, relative error, structural similarity index (SSI), and peak signal-to-noise ratio, as well as clinical relevant indices, such as total perfusion deficit (TPD), were utilized. Overall, AC SPECT images generated using the deep learning networks exhibited good agreement with SPECT CT-AC images, substantially outperforming Chang’s method. The ResNet and UNet models resulted in an ME of −6.99 ± 16.72 and −4.41 ± 11.8 and an SSI of 0.99 ± 0.04 and 0.98 ± 0.05, respectively. Chang’s approach led to ME and SSI of 25.52 ± 33.98 and 0.93 ± 0.09, respectively. Similarly, the clinical evaluation revealed a mean TPD of 12.78 ± 9.22% and 12.57 ± 8.93% for ResNet and UNet models, respectively, compared to 12.84 ± 8.63% obtained from SPECT CT-AC images. Conversely, Chang’s approach led to a mean TPD of 16.68 ± 11.24%. The deep learning AC methods have the potential to achieve reliable AC in MPI-SPECT imaging.

Advanced Quantitative MSK Imaging.

For many years, imaging of the musculoskeletal (MSK) system has focused on subjective and qualitative analysis. However, quantitative assessment of imaging data may provide precise and reproducible image-derived metrics, and through their use there is a promise of increased value of imaging. Specifically, these metrics, or biomarkers, may be used for improved diagnosis, determination of patient prognosis, and monitoring of treatment response. This issue of Seminars in Musculoskeletal Radiology is dedicated to quantitative MSK imaging. Giraudo et al provide a comprehensive overview of the quantitative methods that can be applied to assess inflammatory rheumatologic diseases, considering not only different radiologic techniques but also hybrid imaging. In addition, they describe emerging technologies such as artificial intelligence and radiomics. Burge and Jawetz discuss advanced magnetic resonance imaging (MRI) techniques in osteoarthritis that can comprehensively evaluate joint health, providing additional information regarding both cartilage status and the state of surrounding articular tissues. Gitto and colleagues discuss the currently available state-of-the-art quantitative ultrasound imaging tools available in MSK radiology including shear-wave elastography, perfusion with contrast-enhanced ultrasound and superb microvascular imaging, and quantitative ultrasound assessment of osteosarcopenia. Hemke et al present an overview of quantitative imaging methods to assess the distribution of lean tissue, adipose tissue, and bone in the human body, and their relationship with various pathologies. They stress the importance of body composition imaging for predicting overall health and risk profiling of patients. Jerban and colleagues summarize the quantitative MRI techniques recently developed for cortical and trabecular bone evaluation. Compared with radiograph-based techniques that focus on measuring bone mineral density, the new quantitative MRI techniques can comprehensively evaluate all the components of bone including the organic matrix, bone minerals, water, and fat. Weber et al review the role of neuromuscular imaging in the diagnosis of inherited and acquired muscle diseases. They discuss whole-body imaging as well as more advanced MRI applications including nonproton MRI and multiparametric quantitative imaging that provides deeper insights into muscle pathophysiology. Griffith describes the role of quantitative imaging for the assessment of spine abnormalities. These quantitative techniques have led to a greater understanding of spinal development, marrow physiology, and disease pathogenesis. In their article on quantitative musculoskeletal tumor imaging, Howe and colleagues highlight the emerging role of various quantitative MRI and positron emission tomography (PET) techniques for the characterization, staging, prognosis, and follow-up of bone and soft tissue tumors. They also touch on the role of computed tomography and MRI-based virtual surgical planning and three-dimensional printing as a means to optimize surgical treatment of MSK tumors. Yoon et al review novel approaches to pinpoint MSK pain generators with the use of PET imaging. They also discuss the added value of new radiotracers and of combining information in one hybrid PET/MRI examination. Liu presents the concepts of deep learning and demonstrates how these methods applied to quantitative MRI may accelerate image acquisition and reconstruction, improve tissue segmentation, and aid in the detection of structural knee abnormalities. Finally, Visser and colleagues discuss the trend in medicine that is moving from volume- to value-based payments, and they review the role that quantitative imaging may play. They also discuss the influence of artificial intelligence on value-based health care. The variety of applications across all imaging modalities presented in this issue exemplifies the significant effort being made in developing and validating quantitative imaging techniques applied to the MSK system. The fact that several novel applications are still at the stage of development and optimization while other, more established, methods are already being translated into the clinical applications, demonstrates the tremendous ongoing activity in the field of quantitative MSK imaging. We thank all the authors for their outstanding contribution to this issue. We hope the readers will be as excited as we are to apply quantitative imaging biomarkers in their clinical practice.

Scale effects in remotely sensed greenspace metrics and how to mitigate them for environmental health exposure assessment

Metrics representing exposure to the natural environment are widely used in environmental health-related studies. They are calculated using a variety of different data sources representing greenspace and a range of buffer sizes representing human interaction with the environment. Previous studies have identified issues relating to buffer distance and scaling effects on greenspace exposure assessments when using satellite image-derived metrics. We evaluate the spatial scale sensitivity of three common greenspace metrics (i.e., Normalised Difference Vegetation Index- NDVI, Leaf Area Index- LAI, and Land Use and Land Cover-LULC), using lacunarity analysis, as a scale-dependent measure of heterogeneity based on the principles of fractals. By producing a ‘lacunarity curve’ across multiple spatial scales, we defined the scale-variances for specific greenspace metrics, including the upper scale limit at which the metrics become invariant, approximately 640 m for Sentinel-2 and 480 m for Landsat-8. Each of the greenspace metrics we considered exhibited scale sensitivities, meaning that each is expected to have a different influence on the strength and significance of the statistical associations found between greenspace exposure and health depending on the spatial scale of analysis (e.g., buffer distance). Using lacunarity curves, we produced a novel composite, multi-scale greenspace ‘exposure index’ in which each input scale is weighted according to its relative scale sensitivity. We also created a multi-scale, multi-metric map combining the different vegetation measures while accounting for scale. We found that cumulative exposure gradients across a large urban conurbation are even more marked when using our multi-scale ‘exposure index’ maps compared to traditional approaches. Our multi-scale, composite greenspace ‘exposure index’ mapping techniques are not as vulnerable to scale effects as traditional approaches and can be readily transferred to the analysis of other environmental exposure variables such as air pollution.

Dynamic Mitochondrial Migratory Features Associated with Calcium Responses during T Cell Antigen Recognition.

A T cell clone is able to distinguish Ags in the form of peptide-MHC complexes with high specificity and sensitivity; however, how subtle differences in peptide-MHC structures translate to distinct T cell effector functions remains unknown. We hypothesized that mitochondrial positioning and associated calcium responses play an important role in T cell Ag recognition. We engineered a microfluidic system to precisely manipulate and synchronize a large number of cell-cell pairing events, which provided simultaneous real-time signaling imaging and organelle tracking with temporal precision. In addition, we developed image-derived metrics to quantify calcium response and mitochondria movement. Using myelin proteolipid altered peptide ligands and a hybridoma T cell line derived from a mouse model of experimental autoimmune encephalomyelitis, we observed that Ag potency modulates calcium response at the single-cell level. We further developed a partial least squares regression model, which highlighted mitochondrial positioning as a strong predictor of calcium response. The model revealed T cell mitochondria sharply alter direction within minutes following exposure to agonist peptide Ag, changing from accumulation at the immunological synapse to retrograde movement toward the distal end of the T cell body. By quantifying mitochondria movement as a highly dynamic process with rapidly changing phases, our result reconciles conflicting prior reports of mitochondria positioning during T cell Ag recognition. We envision applying this pipeline of methodology to study cell interactions between other immune cell types to reveal important signaling phenomenon that is inaccessible because of data-limited experimental design.

Assessment of image-derived risk factors for natural course of unruptured cerebral aneurysms.

The goal of this prospective longitudinal study was to test whether image-derived metrics can differentiate unruptured aneurysms that will become unstable (grow and/or rupture) from those that will remain stable. One hundred seventy-eight patients harboring 198 unruptured cerebral aneurysms for whom clinical observation and follow-up with imaging surveillance was recommended at 4 clinical centers were prospectively recruited into this study. Imaging data (predominantly CT angiography) at initial presentation was recorded. Computational geometry was used to estimate numerous metrics of aneurysm morphology that described the size and shape of the aneurysm. The nonlinear, finite element method was used to estimate uniform pressure-induced peak wall tension. Computational fluid dynamics was used to estimate blood flow metrics. The median follow-up period was 645 days. Longitudinal outcome data on these aneurysm patients-whether their aneurysms grew or ruptured (the unstable group) or remained unchanged (the stable group)-was documented based on follow-up at 4 years after the beginning of recruitment. Twenty aneurysms (10.1%) grew, but none ruptured. One hundred forty-nine aneurysms (75.3%) remained stable and 29 (14.6%) were lost to follow-up. None of the metrics-including aneurysm size, nonsphericity index, peak wall tension, and low shear stress area-differentiated the stable from unstable groups with statistical significance. The findings in this highly selected group do not support the hypothesis that image-derived metrics can predict aneurysm growth in patients who have been selected for observation and imaging surveillance. If aneurysm shape is a significant determinant of invasive versus expectant management, selection bias is a key limitation of this study.

Comprehensive assessment of nuclear and cortical backscatter metrics derived from rotating Scheimpflug images

To quantify cataract in lens-backscatter metrics derived from Pentacam Scheimpflug photographs and determine the most predictive metrics for lens grading and visual performance measures. Large tertiary referral hospital in South Australia. Evaluation of diagnostic test or technology. In patients attending a cataract clinic between 2005 and 2008, standardized clinical assessments were performed and Scheimpflug photographs taken. Thirteen measures of cataract backscatter were drawn from the images of nuclear and cortical cataract, including new metrics for cortical cataract. The relationships between lens-scatter metrics, visual performance, and lens grades were determined with stepwise multiple linear regression. The study enrolled 342 eyes (177 patients). The Lens Opacities Classification System III (LOCS III) nuclear opalescence was best predicted by the average backscatter in the anterior half of the nucleus and the whole nucleus average (r(2) = 0.59). The LOCS III cortical grade was best predicted by the number of slices with significant cortical cataract (r(2) = 0.52), as was Pelli-Robson contrast sensitivity (r(2) = 0.16). Visual acuity was best predicted by the combined anterior half and whole nucleus averages (r(2) = 0.20). Pelli-Robson contrast sensitivity under glare was best predicted by a combination of the anterior half and whole nucleus averages and the calculated percentage of cortical cataract (r(2) = 0.16). Scheimpflug image-derived metrics of nuclear scatter were predictive of visual acuity and contrast sensitivity, whereas metrics of cortical cataract were predictive of visual performance under glare. Automated reporting of cataract severity from Scheimpflug images could be useful clinically and should include measurements of cortical cataract. Neither author has a financial or proprietary interest in any material or method mentioned.

Computer-aided detection and diagnosis at the start of the third millennium.

Computer-aided diagnosis has been under development for more than 3 decades. The rate of progress appears exponential, with either recent approval or pending approval for devices focusing on mammography, chest radiographs, and chest CT. Related technologies improve diagnosis for many other types of medical images including virtual colonography, vascular imaging, as well as automated quantitation of image-derived metrics. A variety of techniques are currently employed with success, likely reflecting the variety of imagery used, as well as the variety of tasks. Most areas of medical imaging have had efforts at computer assistance, and some have even received FDA approval and can be reimbursed. We anticipate that the rapid advance of these technologies will continue, and that application will broaden to cover much of medical imaging. Acceptance of, and integration of computer-aided diagnosis technology with the electronic radiology practice is a current challenge. These challenges will be overcome, and we expect that computer-aided diagnosis will be routinely applied to medical images.